AMD is openly talking about Zen 5 pricing as "last gen or lower". Stand by for "Tech Day" in the next few weeks. The current bargain ought to expire when that event takes place.
It's amazing to see so many comments on Zen 5 performance, considering nobody is testing actual parts. A few systems are trickling out but it's not like you can order Zen 5 parts from NewEgg. None the less, someone posted some extremely impressive benchmarks using overclocked, DDR5-8000, RAM. These benchmarks do show the 13900K in an unflattering light so it will be interesting to see what comes of the data in the next week or two. The point being, there is some data which suggests AMD is sandbagging Zen 5 performance. With massive latency improvement over Zen 4, turning up the RAM speed could provide out weighted benefit. Time will tell.
I'm pretty confident the Zen 5 retail release will be July 31. That will be the day the anti-AMD BS stops and everyone reverts to pretending they predicted correctly all along. The amount of total crap is mind blowing. Here are some thoughts I consider facts. - Intel uses way more power under load than AMD. For the same job, Intel is not in the same league by perhaps 50% more power required per job and this lead will extend with Zen 5. - AMD Zen 4 uses way more idle power than Intel 13000/14000 by nearly double. Intel 13900 idles at 12~15 Watts. AMD 7950X idles at 22~25 Watts. Computers idle nearly all of their lives so this is huge. We don't yet know if AMD improved this for Zen 5. - Performance between Zen 4 and 13000/14000 is extremely similar in just about every benchmark. Zen 5 will change this such that a few benchmarks will pull way ahead but many benchmarks will be improved but still extremely similar. From what I can tell, even AVX512 instructions do not double in performance, despite doubling the buss bandwidth in Zen 5 over Zen 4. AMD slides show 45% boost in AVX512 over Zen 4. - Intel has an x/2 clock multiplier which allows them to dominate when memory is overclocked heavily. We don't know if Zen 5 adds x/2 but, for now,Intel is the only choice for heavy overclockers. - Zen 5c will be released at 3nm toward the end of this year. It should bring direct competition to Lunar Lake for laptops. The narrative that Intel will dominate laptops is crap. The correct narrative is: time will tell who will dominate the laptop performance benchmarks but it will not be Snapdragon X Elite, except for battery efficiency. - As best I understand, x86_64 is a bit more efficient than ARM at work loads. ARM basted CPUs tend to be heavily more efficient than x86_64 at idle and light work loads. ARM will and should sweep the tablet form factor but I expect it to get zero traction in the gaming world.
Like x1000. I could not agree more with this analysis. As far as I know, AMD GPU idle draw is still too high as well. My XFX 6950xt fans spin intermittently doing nothing. Pretty aggressive default fan profile as the temps are not that high, but still, it is an issue going back a few generations. I would have gone Nvidia again, but damned if I pay those prices for raster performance. The price I paid for my card was a no-brainer. I do miss CUDA though.
AMD has made a lot of improvements to their GPU. RDNA 4 will bring a dual chiplet design. I haven't run an Nvidia GPU for quite some time. Nvidia is notnlinux friendly (at one time, they were the most Linux friendly). Linux is gaining popularity but not by any amount that is relevant. We are seeing Windows refugees start to shuffle in. The real question is, what will happen in on January one when Windows 10 goes end of security support. We have two Windows laptops and I still don't know what we are going to do. Probably I will go 100% Linux and my wife will get a new Windows laptop with the latest MS cloudware. I'm surprised the MS cloud strategy is working as well as it is but it appears MS will successfully bring the corporate world into their cloud with extreme few exceptions. All of that bodes well for Intel.
Rapidfire update: - Intel is going to delay Lunar Lake a bit, almost certainly to bring more game to the table to compete against Zen 5. This will provide both clear air to AMD, as well as likely limit how much laptop share AMD can achieve on the short term. - There is talk from a very credible web site that Zen 5 X3D variants will use the same cache chiplet as Zen 4 X3D currently does. This is in direct conflict with several other rumors and I hope it's not true. The current X3D cache does not cache the entire CPU complex. I believe it will only work on one chiplet, so half of the cores, but is tuned for one or two cores under heavy load. There are a ton of rumors the new X3D will allow the cache to optimize all cores. It's possible the new X3D variant is not ready and they will be shipping 9000X3D with the old cache chiplet. If true, this is a big hit. - Intel had a rough time with power/clock issues on the 13900/14900 processors. I will mention the 7000X3D had quite a few failed parts, when first introduced, due to allowing unlimited over-volting. Limiting the voltage clipped over-clocker's wings but also ended the part failures. Intel continues to struggle with their platform. It is very close to stable with specific settings but Intel is flying extremely close to the sun. Gotta say, advantage AMD on this issue... - There is so much AMD misrepresentation and lies, much of which seems blatantly deliberate, any investor should ignore all but trusted news sources. - We are seeing some hum-drum 9950X3D benchmark figures floating around leakville but those benchmarks were done with the CPU limited to 60 Watts. This configuration prevents all of the boost mechanisms from engaging, or even achieving full base clock. Despite the handicap, the 9950X3D still scores at the very top of the benchark list or second place, in some situations. I expect 30% more performance, once the part can be tested under full power with strong cooling to achieve maximum boost clock. - It isn't surprising there are so many lies and wildly ignorant conclusions floating around. What I find myself surprised at is how much of this crap is being picked up by major media outlets. I'm still scratching my head that Bloomberg picked up the 40% IPC uplift crap that every single person who is reasonably knowledgeable about computers knew was nonsense. The current flurry of FUD is being picked up widely, also. Apparently, they could care less if they are proven to be complete morons in three weeks time.
I just want to comment on the X3D step situation because I think it matters but probably doesn't. In current form, X3D is only helpful to most games and almost nothing else. There are probably general compute tasks which benefit from X3D but these aren't being identified by productivity software benchmarks. A more generalized X3D cache would allow highly multi-processing applications to see a nice boost. Compute heavy tasks like video encoding/transcoding would definitely notice it but the advantage needs to be observed on all cores. Why I think this matters: Right now, X3D chips are ONLY of interest to some gamers. Even there, most games are GPU bound, not CPU bound. If X3D were to provide any noticeable improvement in general compute, even if only a couple of percent, AMD could massively increase sales volume of these ultra-high end parts. I am 90% sure, the 9000X3D will ship with a new step cache chiplet but we won't know for sure until the parts are tested by a bunch of independent nerds with an anime fetish.
One more reason why my next system will have an AMD CPU: Being quiet this long with such erratic error data whether it be enthusiast or server MBs suggests to me that there is something really wrong with the architecture of the 13th and 14th gen Intel CPUs. The top SKUs from both gens are at the bleeding edge and show it the most naturally. I can only imagine what the big dogs like Dell, HP, and Lenovo are thinking right now.
I'm not so sure a place like Bloomberg has a token nerd with an anime fetish on their payroll to vet the quality of such statements. In fact, the more outlandish the claim, the more clicks they get. The random person who reads it is not going to check the accuracy of the reporting after the fact. Only nerds like you and me would do such a thing
I can't help but notice a 15% performance uplift is being widely reported for the 9950X. This is undoubtedly based on expanding benchmarks from engineering samples in the wild. If these morons plagiarizing other site headlines had read even a couple of sentences past the headline before copy/pasting it on their own site, they might have noticed the 15% uplift comes at 160W pumping into an engineering sample. These are 170W CPUs so there is a tiny bit to be gained with TDP tuning and production CPUs are said to be good for a couple of hundred extra MHz. That will put the 9950X comfortably 16% above the 7900X, making AMD's performance claim just slightly conservative. Why do people follow and respond to sources which have proven to be wrong the vast majority of the time? They specialize in wrong, and yet people continue to follow? The trading community does the same thing. Someone could have gotten the last 100 predictions wrong and yet masses will tune in to hear what the next prediction is and enough will do it to cause a market response to the news. Meanwhile, a couple of dudes from Stockaholics are dead accurate by curating a couple of trusted sources and a small amount of reading.
I think we see more consolidation in the $140-$150 area, but I still like it. I was looking at getting a new laptop and a lot of the nice options were all AMD CPUs.
I expect the trend toward AMD CPUs will continue on the short term, as Intel has a small quagmire with CPU degredation on their top models. Lower model Intel CPUs continue to be desireable and it will take considerable time for AMD to gain laptop market share. The real estimate here is how much laptop market share AMD can carve out of Intel. AMD currently holds 21% of laptop market share. I think they will be lucky to get to 50% market share by the time Intel responds with their next generation platform. Assuming the next generation Intel platform arrives in volume for 1Q25 (they will paper release it later this year but volume will take a couple of months, minimum) and assuming the next generation Intel CPU proves desirable, AMD market share gains will begin to recede at the start of 2025. I think Zen 6 will be a nice step forward for AMD but not a revolution, by any means. Perhaps it will re-take the lead from Intel when Zen hits the streets in the second half of 2025 (best case). Qualcomm seems to be less of a threat than initially thought. AI also seems to be less of a must-have feature for the vast majority than AMD thought. Place your bets....
AMD delays Ryzen 9000 series to mid August due to unspecified quality issues. https://www.anandtech.com/show/2148...aunch-1-to-2-weeks-due-to-chip-quality-issues
AMD looks like a better value at around $135-$140 than it did at $170+. Added to my watchlist. I've had some in the past, but took the money when it seemed like it was topping out. I am mostly a long term, buy-and-hold investor, but there are too many opportunities to 'buy low' and 'sell high' to pass them all up. It's not timing the market, it seems to me to be common sense. Especially with no fee trades and trailing stop loss orders - not much downside! I earned my wings back when trading fees were relatively high and the amount of data available and the speed at which it could be obtained were much lower. People seem to forget that it wasn't that long ago that $10 trading fees for stocks was considered a LOW fee! $40 was the norm for a while. I went back and looked it up - NO FEE TRADES have only been commonplace for about 5 or 6 years!
AMD news is all over the map. There are some severe reviews, trashing Zen 5, and showing benchmarks with 0% gain over Zen 4. There are also reviews showing 13~24% IPC uplift. The 0% gain crowd did their testing with 65 Watt PBO in place. lol! This storm will blow through and Zen 5 is going to be just fine. I'm staggered how many people are prepared to lie about AMD to smear AMD when Intel is in trouble. We live in a post truth world.
